[PATCH] m68k: cmpxchg: Dereference matching size

[PATCH] m68k: cmpxchg: Dereference matching size

[Kees Cook]

Similar to the recent arm64 fix[1], avoid overly wide casts in the cmpxchg
implementation. Avoid this warning under -Warray-bounds with GCC 11:

net/sched/cls_tcindex.c: In function 'tcindex_set_parms':
./arch/m68k/include/asm/cmpxchg.h:64:17: warning: array subscript 'volatile struct __xchg_dummy[0]' is partly outside array bounds of 'struct tcf_result[1]' [-Warray-bounds]
   64 |                 __asm__ __volatile__
      |                 ^~~~~~~
net/sched/cls_tcindex.c:338:27: note: while referencing 'cr'
  338 |         struct tcf_result cr = {};
      |                           ^~

No binary output differences are seen from this change.

[1] commit 3364c6ce23c6 ("arm64: atomics: lse: Dereference matching size")

Signed-off-by: Kees Cook <keescook@chromium.org>
---
 arch/m68k/include/asm/cmpxchg.h | 9 ++++-----
 1 file changed, 4 insertions(+), 5 deletions(-)

diff --git a/arch/m68k/include/asm/cmpxchg.h b/arch/m68k/include/asm/cmpxchg.h
index e8ca4b0ccefa..6cf464cdab06 100644
--- a/arch/m68k/include/asm/cmpxchg.h
+++ b/arch/m68k/include/asm/cmpxchg.h
@@ -4,8 +4,7 @@
 
 #include <linux/irqflags.h>
 
-struct __xchg_dummy { unsigned long a[100]; };
-#define __xg(x) ((volatile struct __xchg_dummy *)(x))
+#define __xg(type, x) ((volatile type *)(x))
 
 extern unsigned long __invalid_xchg_size(unsigned long, volatile void *, int);
 
@@ -50,7 +49,7 @@ static inline unsigned long __xchg(unsigned long x, volatile void * ptr, int siz
 			 "1:\n\t"
 			 "casb %0,%1,%2\n\t"
 			 "jne 1b"
-			 : "=&d" (x) : "d" (x), "m" (*__xg(ptr)) : "memory");
+			 : "=&d" (x) : "d" (x), "m" (*__xg(u8, ptr)) : "memory");
 		break;
 	case 2:
 		__asm__ __volatile__
@@ -58,7 +57,7 @@ static inline unsigned long __xchg(unsigned long x, volatile void * ptr, int siz
 			 "1:\n\t"
 			 "casw %0,%1,%2\n\t"
 			 "jne 1b"
-			 : "=&d" (x) : "d" (x), "m" (*__xg(ptr)) : "memory");
+			 : "=&d" (x) : "d" (x), "m" (*__xg(u16, ptr)) : "memory");
 		break;
 	case 4:
 		__asm__ __volatile__
@@ -66,7 +65,7 @@ static inline unsigned long __xchg(unsigned long x, volatile void * ptr, int siz
 			 "1:\n\t"
 			 "casl %0,%1,%2\n\t"
 			 "jne 1b"
-			 : "=&d" (x) : "d" (x), "m" (*__xg(ptr)) : "memory");
+			 : "=&d" (x) : "d" (x), "m" (*__xg(u32, ptr)) : "memory");
 		break;
 	default:
 		x = __invalid_xchg_size(x, ptr, size);
-- 
2.30.2

Re: [PATCH] m68k: cmpxchg: Dereference matching size

[Geert Uytterhoeven]

Hi Kees,

On Wed, Feb 2, 2022 at 11:53 PM Kees Cook <keescook@chromium.org> wrote:
> Similar to the recent arm64 fix[1], avoid overly wide casts in the cmpxchg
> implementation. Avoid this warning under -Warray-bounds with GCC 11:
>
> net/sched/cls_tcindex.c: In function 'tcindex_set_parms':
> ./arch/m68k/include/asm/cmpxchg.h:64:17: warning: array subscript 'volatile struct __xchg_dummy[0]' is partly outside array bounds of 'struct tcf_result[1]' [-Warray-bounds]
>    64 |                 __asm__ __volatile__
>       |                 ^~~~~~~
> net/sched/cls_tcindex.c:338:27: note: while referencing 'cr'
>   338 |         struct tcf_result cr = {};
>       |                           ^~
>
> No binary output differences are seen from this change.
>
> [1] commit 3364c6ce23c6 ("arm64: atomics: lse: Dereference matching size")
>
> Signed-off-by: Kees Cook <keescook@chromium.org>

Thanks for your patch!

Reviewed-by: Geert Uytterhoeven <geert@linux-m68k.org>

Do you want me to queue this in the m68k for-v5.18 branch, or do
you want to take it yourself, together with commit a8712a32665f9b1a
("Makefile: Enable -Warray-bounds") in linux-next that triggers this?
In case of the latter:
Acked-by: Geert Uytterhoeven <geert@linux-m68k.org>

Please let me know. Thanks!

Gr{oetje,eeting}s,

                        Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds

Re: [PATCH] m68k: cmpxchg: Dereference matching size

[Kees Cook]

On Thu, Feb 03, 2022 at 09:13:30AM +0100, Geert Uytterhoeven wrote:
> Hi Kees,
> 
> On Wed, Feb 2, 2022 at 11:53 PM Kees Cook <keescook@chromium.org> wrote:
> > Similar to the recent arm64 fix[1], avoid overly wide casts in the cmpxchg
> > implementation. Avoid this warning under -Warray-bounds with GCC 11:
> >
> > net/sched/cls_tcindex.c: In function 'tcindex_set_parms':
> > ./arch/m68k/include/asm/cmpxchg.h:64:17: warning: array subscript 'volatile struct __xchg_dummy[0]' is partly outside array bounds of 'struct tcf_result[1]' [-Warray-bounds]
> >    64 |                 __asm__ __volatile__
> >       |                 ^~~~~~~
> > net/sched/cls_tcindex.c:338:27: note: while referencing 'cr'
> >   338 |         struct tcf_result cr = {};
> >       |                           ^~
> >
> > No binary output differences are seen from this change.
> >
> > [1] commit 3364c6ce23c6 ("arm64: atomics: lse: Dereference matching size")
> >
> > Signed-off-by: Kees Cook <keescook@chromium.org>
> 
> Thanks for your patch!
> 
> Reviewed-by: Geert Uytterhoeven <geert@linux-m68k.org>
> 
> Do you want me to queue this in the m68k for-v5.18 branch, or do
> you want to take it yourself, together with commit a8712a32665f9b1a
> ("Makefile: Enable -Warray-bounds") in linux-next that triggers this?
> In case of the latter:
> Acked-by: Geert Uytterhoeven <geert@linux-m68k.org>

Given how -next is being ordered, it's probably best to go via my tree.
Thanks for the review!

-Kees

> 
> Please let me know. Thanks!
> 
> Gr{oetje,eeting}s,
> 
>                         Geert
> 
> --
> Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org
> 
> In personal conversations with technical people, I call myself a hacker. But
> when I'm talking to journalists I just say "programmer" or something like that.
>                                 -- Linus Torvalds

-- 
Kees Cook